K 10
svn:author
V 5
andre
K 8
svn:date
V 27
2012-11-12T09:50:26.994756Z
K 7
svn:log
V 653
Continue conversion of bge(4) into a hybrid interrupt/polling
model including live-lock prevention.

Simplify bge_setup_tso() based on stack guarantees on minimal
packet layout.

Rename bge_rxeof() to bge_rx() to better describe its purpose.

Merge bge_ithr_msix() and bge_ithr() together as bge_ithr().
Change the order of DMA ring processing.  First do the tx
completion ring to free up space and allow sending to go on.
Then dequeue all packets from the rx ring, drop the lock and
send them up the stack.  Do a maybe_yield() to observe the
available quantum and let other threads run in between.

Remove bge_poll() as unnecessary.

Work in progress.

END